How to teach your parent to use ChatGPT: a 30-minute setup conversation, scripted
The thirty-minute setup conversation with a parent covers six things in order: signing in, asking the first question, what to do when ChatGPT is wrong, the five things to never type in, how to get unstuck when you aren't there, and where to write the URL so they can find it tomorrow. The trick is to say the actual sentences out loud rather than handing them a list.
The reason this piece exists is that I have now watched, in person and over the phone, about a dozen of these conversations. Some of them went well. The ones that went well had almost nothing in common with each other except that the adult child sitting next to the parent had decided, before sitting down, what they were going to say. Not what they were going to cover — what they were going to literally say, out loud, in plain English, while the parent watched the screen.
That is what this piece is. It is the script.
It is written for you — the adult child, somewhere between forty-five and sixty, who has either bought your mother or father a ChatGPT setup or has signed them up on your laptop and is about to hand it over. According to the Pew Research Center, about seventy-three million Americans are now informally caring for an aging family member in some capacity, and a significant share of that care now includes tech support. You are one of those seventy-three million. The thirty minutes that follow are part of the job.
Before you sit down: three things to have ready
One: a sign-in already created. Do not create the account in front of them. Account creation involves CAPTCHAs, phone verification, and at least one screen that tries to upsell them to the paid version. Do that part yourself at the kitchen table the night before. (If you haven't yet, I wrote a separate piece on why the free version is the right one to start on, and how to keep them on it.)
Two: a pen and an index card. They are going to want to write the URL down. Let them. The index card is more important than anything on the screen.
Three: thirty minutes that are not interrupted. Not a phone in the kitchen. Not a grandchild coming in to ask a question. The whole point is that they get to be the slow person in the room for half an hour, which means you have to clear the room.
Now sit down. They sit in front of the screen. You sit beside them, not behind them. Behind them is the posture of someone watching them get tested. Beside them is the posture of someone helping them figure something out together. That distinction is not soft. It is the entire emotional frame of the next half hour.
Minute 1 to 5: signing in
Open the browser. Go to chatgpt.com. The page loads. There is a box. Below the box, there are some example questions.
Here is what you say. Out loud. In this order.
"Okay, this is it. This is the whole thing. There is one box in the middle of the page. You type a question into the box. It types an answer back. That is the entire app. Everything else is just a setting somewhere."
That sentence does more work than any other sentence in the thirty minutes. It tells them, before they ask, that they are not missing anything. There are not three other features they have to learn about. There is the box. That is it.
Then click "Log in" and sign them in with the account you created. The first time, you may have to do this part. The second time, you point at where the email field is and let them type it. Let them type it. Their hands need to remember the keys, not yours.
Once they're logged in, you say the second thing. "This page is called the home page. The button on the top left says 'New chat.' Every time you want to start a new question that has nothing to do with the last one, you click that. We're not going to click it yet."
That's minute five.
Minute 5 to 10: the first question
This is the part most people get wrong. They tell the parent to "try it out" and the parent freezes. The parent freezes because they have not used a tool in fifty years that responds to questions in plain English, and they have no idea what register to use. Should I be formal? Should I say please? Should I sound like a search engine?
You solve this by giving them the first question. The first question is not theirs. It is yours, and you give it to them as a script.
"I want you to type exactly this. Don't worry about it being a good question. We're just going to see what happens."
Then you say:
"Write me a thank-you note to my neighbor for bringing over a casserole last week. Make it warm but not too long."
They type it. Slowly. They are still hunting for keys. That's fine. Don't reach over and type for them. Let them finish the sentence. When they hit Enter, ChatGPT writes a thank-you note in about four seconds.
Now the third thing you say. "See what it did? It wrote you a draft. Not the final thing. A draft. You can copy it, you can change it, you can ask it to make it shorter, you can ignore it. The point is you didn't have to start from a blank page. That is what this tool does — it gets you off the blank page."
If they laugh, you are halfway home. Most parents laugh at the first thank-you note, because the model produces something slightly more effusive than they would ever write, and the gap between how I sound and how this thing thinks I sound is genuinely funny. That laugh is the moment they decide the tool is not scary.
Minute 10 to 15: what to do when ChatGPT is wrong
This is the part that most adult children skip and then regret. You have to tell them, out loud, before they encounter it on their own, that the tool gets things wrong sometimes.
Here is the sentence. "This thing is going to be wrong sometimes. Not in a scary way. In the way that a confident person at a dinner party is sometimes wrong about a date. For writing letters and rephrasing things, it almost never matters. For looking up a phone number or a medical dose or anything where being exactly right matters, you have to check it against a real source."
Then you show them the regenerate button — the little arrow under the response that asks ChatGPT to try again. The OpenAI Help Center documents the regenerate function in its main user guide and it does what it says: it asks the model to write a different version. Show them how to click it. Show them that the new answer might be better or worse. Show them that they get to pick.
A 2023 study published in the Journal of Medical Internet Research on technology adoption among older adults found that the single biggest predictor of continued use was not the parent's confidence on day one — it was whether they had someone they could ask when the tool surprised them in the first two weeks. The "surprise" is almost always either it gave me a weird answer or I clicked something I didn't mean to. Both are survivable if you have told them, before it happens, that the tool sometimes does that.
Minute 15 to 20: the five things never to type in
This is the safety part. You can read it almost verbatim. You can also just hand them the card from the longer safety piece I wrote, but it is better if they hear it in your voice first.
"There are five things you should never paste into this box. Your Social Security number. Your bank or credit card numbers. Any password to anything. Your full medical records with your name and date of birth on them. And anyone's full name combined with their address or school, especially the grandkids."
"Nothing else is on the list. Recipes, letters, questions about your medication in general, things from the newspaper, family stories — all of that is fine. The five things above are the line."
Then you write the five things on the index card. Big block letters. You leave the card next to the keyboard. Not in a drawer. Next to the keyboard.
The AARP Foundation's Senior Planet curriculum — which has been teaching older adults to use the internet for more than two decades — uses the same rule: write the rule by hand, leave it visible, never make them remember it. The card is the trick. The card outlives the conversation.
Sidebar: Common things parents will ask in the first hour (and how to answer) "Where did the answer go?" — Scroll up. Old messages stay on the screen. Nothing is deleted. "How do I start over with a new question?" — Click "New chat" on the top left. The old chat is saved on the left-hand list. It is not gone. "Is this thing listening to me?" — No. ChatGPT only sees what you type. The microphone is a separate button you have to press, and you don't need it. "Can other people see what I asked?" — No. The conversation is between the account and OpenAI. Nobody you know can read it. "Why is it pretending to be a person?" — That's the style the company picked. It's a writing tool that writes in first person. You can ignore the "I" and just read the words. "Did I break it?" — No. Whatever you did, you didn't break it. Refresh the page. Try again.
Minute 20 to 25: how to get help when you aren't there
This is the most important five minutes, and it is the one most people forget. You will not be there next time. They need a plan for what to do when the screen does something they don't recognize.
Here is what you do. You write three things on the second side of the index card.
- The URL: chatgpt.com
- Your phone number, labeled "if I'm stuck, call me first"
- The Plain English Company support email, labeled "or write here, they answer in plain English" — or, if they have a different setup binder, the equivalent contact from whoever made it.
Then you say the last big thing.
"If something on the screen doesn't make sense, the first thing to do is close the tab and reopen it. About half the problems fix themselves. If it still doesn't make sense, call me. If I don't pick up, write to the email on the card. Do not click anything that says you need to pay to fix the problem. There is no such message in this tool. If you see one, it is a scam, and you should close the tab and call me."
That last sentence is the one that matters. The reason it matters is that scam pop-ups in browsers are common, and a parent who has just been told this is a new tool you don't fully understand is exactly the person who clicks a pop-up that says their computer is infected. Tell them, before they see one, that the answer is always to close the tab and call you. The card is the backstop.
Sidebar: What to do when your parent says "this doesn't make sense" They will say this. Probably in the first week. Sometimes in the first hour. The instinct is to take the keyboard from them and fix whatever is on the screen. Don't. Instead, three steps. One: ask them what they typed, in their own words. Two: ask them what they expected the answer to be. Three: read the answer the model gave, out loud, together. Nine times in ten, the problem is not technical. The problem is that the model wrote a fine answer to a question that wasn't the one your parent meant to ask. The fix is to type the question again, slightly differently. Show them the regenerate arrow. Show them that the answer they got is not the only one available. If they are getting frustrated, stop. Close the laptop. Make a cup of coffee. The thing they need to learn is not the tool. It is the feeling that they can close the laptop and the tool will still be there tomorrow, exactly as they left it. The OpenAI Help Center documents that chat history is saved automatically on every account by default — they can pick up the same conversation tomorrow morning. That fact, said out loud, often does more to calm them than any tip.
Minute 25 to 30: the goodbye
The last five minutes are the ones that decide whether they ever open the tab again after you leave.
Have them do one more question on their own. You don't pick the question. Let them pick. If they pick something hard — "help me write the eulogy for Uncle Bill" — let them. ChatGPT is unusually good at that kind of writing. If they pick something silly — "what is a sonnet" — that's fine too. The point is that the last action of the conversation belongs to them, not you.
Then, before you stand up, write the URL one more time. On the front of the index card. In the largest letters that fit. Tape the card to the desk if there is tape in the drawer. If there isn't, put the card under the keyboard. They will find it tomorrow morning.
Tell them: "You can't break it. You can close the tab and come back any time. Everything you typed today is still there if you want to look at it. I'll check on you in a couple of days."
Then leave. Don't linger. Lingering communicates that you are worried they can't do this without you, which is the opposite of the message they need to walk away with.
The reason the binder I make exists is that this thirty-minute conversation is not a thing every adult child has time to do, and it is not a thing every parent wants to do with their adult child in the room. Some parents would rather read a printed book at their own pace, on their own time, with no one watching. For those parents — and for the gift-givers who would rather hand over an object than schedule a session — the binder is the physical reference that lives next to the computer in place of you. It is $19 on the pricing page, thirty pages, plain English, spiral-bound so it lies flat next to the keyboard. The one-page free version of the first-day plan, if you want to test the voice before you buy the longer one, is at the start link. If you are still earlier in the decision and trying to figure out what to actually buy a parent who isn't tech-savvy, that piece is the one to read first.
— Shawn Ivie Founder, Plain English Company